Revert JIT VmClosure optimization — was producing wrong bytecode

The optimization to call the compiler through VM directly (instead
of CEK) when it's JIT-compiled was producing incorrect bytecode for
all subsequently compiled functions, causing "Expected number, got
symbol" errors across render-to-html, parse-loop, etc.

Revert to always using CEK for compilation. The compiler runs via
CEK which is slower but produces correct bytecode. JIT-compiled
USER functions still run at VM speed.

1166 passed, 0 failed.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

hosts/ocaml/lib/sx_vm.ml

@@ -575,13 +575,7 @@ let jit_compile_lambda (l : lambda) globals =
     let param_syms = List (List.map (fun s -> Symbol s) l.l_params) in
     let fn_expr = List [Symbol "fn"; param_syms; l.l_body] in
     let quoted = List [Symbol "quote"; fn_expr] in
-    let result = match compile_fn with
-      | VmClosure cl ->
-        (* Compiler loaded as bytecode — call through VM directly *)
-        call_closure cl [quoted] globals
-      | _ ->
-        (* Compiler loaded from source — call through CEK *)
-        Sx_ref.eval_expr (List [compile_fn; quoted]) (Env (make_env ())) in
+    let result = Sx_ref.eval_expr (List [compile_fn; quoted]) (Env (make_env ())) in
     (* Inject closure bindings into globals so GLOBAL_GET can find them.
        Only injects values not already present in globals (preserves
        existing defines). Mutable closure vars get stale snapshots here